Hsr Has Completely Broken Hit Detection
#21
Posted 16 June 2013 - 12:39 PM
I legged a Spider, and then proceeded to pound it with my Atlas RS with 4 Large Pulses and an AC20, it took easily a total of 3 60dmg alphas all over his face, and yet managed to survive through that.
I did get the kill in the end, but holy **** wtf... It's even worse than the uncorable Raven pre-nerf.
#22
Posted 16 June 2013 - 03:17 PM
w0rm, on 15 June 2013 - 01:37 PM, said:
PGI secretly implemented a hidden D6 so everytime you shoot and score a hit, the system rolls a virtual D6 and determines if any damage is caused or if your shot only ¨glances¨ the target and does decreased to no damage.
#23
Posted 17 June 2013 - 05:33 AM
I had a laughable fight with a HBK-4P against a jenner and spider last night as well. Usually, legging lights with my 4P is my primary goal in life most games, and I am pretty good at it. This time, I would have been better off shining a laser pointer in their eyes for all the damage that landed. I actually gave up on the legs and just started going for torsos once they blew one of my arms off, but that didn't seem to help much either. I know I got 7 ML's on the jenner for a full burn two or three times, and he didn't have a single section of armor that was even dark yellow when I finally died. Usually, if I keep 35 points of lasers the front of a jenner, SOMETHING turns red. I did that at least twice to the same mech and no armor section was even half gone. That surprised me.
Edited by Maxx Blue, 17 June 2013 - 05:35 AM.
#24
Posted 17 June 2013 - 09:51 AM
#26
Posted 17 June 2013 - 11:10 AM
#27
Posted 17 June 2013 - 11:39 AM
I believe basically this is related to the hot fix for the disconnect issue last patch. Which thankfully I can play now but I think everyone is suffering now with hit detection issues.
#28
Posted 17 June 2013 - 12:01 PM
If you get out of cover, as soon as you are out you shoot (moving the crosshair on the target AND as soon as it's on it, click to unleash the weapons) you'll see no hit. This because HSR rewinds your shooting position, few instants before the position that you saw in your client (let's say, your ping + some other time), this makes the server think you shot the cover in front of you, since HSR rolls you back when you were still getting the clearance for your weapons.
The same happens when aiming, if you make last-time adjustments to a target; let's say you aim at the CT, but as soon as the paperdoll updates you see that the RT is critical and you swing your crosshair to hit the RT, notice how the damage is dealt to the CT and not the RT that has been graphically hit. This because HSR rolls your aiming position and physical mech position to an earlier moment.
Learn to pop out of cover, take aim to a fixed spot for a while (LEAD your targets always unless they are nearly static) and shoot. If the target is static **** your own mind by placing the crosshair on it and hold fire for a few instants before shooting, the shots will get where they are supposed to go.
This has been happening since 16th april (ballistic HSR patch); before lasers were working fine, after 16th of april even them have the same issue, shoot them on a moving target correctly hitting it by keeping the crosshair and the beams on it and few damage will be addressed, because HSR will roll you back to an older position that what you saw in your client.
The higher your ping is, the greater is the lead you require. With the last patch things went fubar furthermore, pings have raised a lot and with huge variability (I was stable between 119/126 ping, now it fluctuates continuosly from 130 to even 200) worsening hit detection and HSR unreliability.
Keep hammering the support and bumping threads like these to let them know..
#29
Posted 17 June 2013 - 12:07 PM
soarra, on 06 June 2013 - 10:38 AM, said:
When an AC/20 is detected to hit between the side and center torsos with no favoring between one way or the other, the AC/20 does zero damage. It has less to do with detection and more to do with the person aiming. This is especially noticeable against Spiders because, simply put, they're the easiest ones to hit in between the CT and STs.
#30
Posted 17 June 2013 - 07:06 PM
Koniving, on 17 June 2013 - 12:07 PM, said:
I had to re-read this a number of times to make sure I'm not crazy.
So when the mech is hit, because the server can't determine if it was to the CT or ST, decides it's NEITHER?
How is that a failure on the shooter's part?
If this is true, this is very bad.
#31
Posted 17 June 2013 - 07:35 PM
Haitchpeasauce, on 17 June 2013 - 07:06 PM, said:
So when the mech is hit, because the server can't determine if it was to the CT or ST, decides it's NEITHER?
How is that a failure on the shooter's part?
If this is true, this is very bad.
Well, yes. You're aiming with the assumption that your ballistics will hit one torso or the other, but then they divide up and slam in between the torsos. It hasn't got anything to do with the HSR or detection. Although it does have something to do with a script in the game that isn't set to counter for this.
#32
Posted 17 June 2013 - 07:41 PM
#33
Posted 17 June 2013 - 08:38 PM
Koniving, on 17 June 2013 - 07:35 PM, said:
Is this something that PGI has documented?
Or something you've come across in your testing?
I should spend some time in the testing grounds ...
Edited by Haitchpeasauce, 17 June 2013 - 08:39 PM.
#34
Posted 17 June 2013 - 08:56 PM
#35
Posted 17 June 2013 - 08:56 PM
Spiders, jenners, commando, etc, other fast mechs: I need to do some serious leading. Almost entire widths of a mech infront, sometimes more.
Cataphracts,etc, other slow mechs: I need to aim at at the CT to hit the RT, or the CT to hit the LT, depends entirely on their movement direction, but it's pretty much a torso out of alignment.
Mostly noticed with Autocanons, only way around it is to waste ammo as tracers to guess where the server is putting them and then aim at the spot. Having great difficulties with PPC's registering on fast mechs. It hits dead on, but nothing registers, and I get people in team chat saying "Lead the target noob".
#36
Posted 18 June 2013 - 03:21 AM
John MatriX82, on 17 June 2013 - 12:01 PM, said:
If you get out of cover, as soon as you are out you shoot (moving the crosshair on the target AND as soon as it's on it, click to unleash the weapons) you'll see no hit. This because HSR rewinds your shooting position, few instants before the position that you saw in your client (let's say, your ping + some other time), this makes the server think you shot the cover in front of you, since HSR rolls you back when you were still getting the clearance for your weapons.
The same happens when aiming, if you make last-time adjustments to a target; let's say you aim at the CT, but as soon as the paperdoll updates you see that the RT is critical and you swing your crosshair to hit the RT, notice how the damage is dealt to the CT and not the RT that has been graphically hit. This because HSR rolls your aiming position and physical mech position to an earlier moment.
Learn to pop out of cover, take aim to a fixed spot for a while (LEAD your targets always unless they are nearly static) and shoot. If the target is static **** your own mind by placing the crosshair on it and hold fire for a few instants before shooting, the shots will get where they are supposed to go.
This has been happening since 16th april (ballistic HSR patch); before lasers were working fine, after 16th of april even them have the same issue, shoot them on a moving target correctly hitting it by keeping the crosshair and the beams on it and few damage will be addressed, because HSR will roll you back to an older position that what you saw in your client.
The higher your ping is, the greater is the lead you require. With the last patch things went fubar furthermore, pings have raised a lot and with huge variability (I was stable between 119/126 ping, now it fluctuates continuosly from 130 to even 200) worsening hit detection and HSR unreliability.
Keep hammering the support and bumping threads like these to let them know..
Thank you very much, John. Concise write up and matches my observations. All in all the gameplay feels a bit more synthetic. I don't shoot with intuitive lead, but have to remind myself how my gameplay is affected by the server/client interaction. As with all electronic assist systems it gets much harder if you don't exactly know how it works.
Good old classic lag modem play with ping 250 was just much more predictable and linear. True leading for latency+ projectile velocity.
Edited by Slanski, 18 June 2013 - 03:21 AM.
#37
Posted 18 June 2013 - 03:25 AM
Haitchpeasauce, on 17 June 2013 - 07:06 PM, said:
So when the mech is hit, because the server can't determine if it was to the CT or ST, decides it's NEITHER?
How is that a failure on the shooter's part?
If this is true, this is very bad.
Happens also with Gauss rifles and PPCs, but it's most notable on gauss rifles and can be seen also on testing grounds, on the cataphract and the centurion it's pretty easy to slip the shot between the joint of a side torso and the CT resulting in no damage although sometimes the crosshair becomes red since the hit is registered.
Slanski, on 18 June 2013 - 03:21 AM, said:
Thank you very much, John. Concise write up and matches my observations. All in all the gameplay feels a bit more synthetic. I don't shoot with intuitive lead, but have to remind myself how my gameplay is affected by the server/client interaction. As with all electronic assist systems it gets much harder if you don't exactly know how it works.
Good old classic lag modem play with ping 250 was just much more predictable and linear. True leading for latency+ projectile velocity.
Trust me if I say that I fell in depression after 16th's April patch. For nearly 3 weeks It was like not being able to play the game anymore, I didn't shot anything anywhere, didn't develop damage.. this until I realized the "positional desynchronization" issue due to HSR kicking in.
I was on frozen, and correctly aimed at the head of an highlander, while cornering from a palace. All my 4 PPCs entered the cockpit graphically and the hit was awarded, but the paperdoll updated itself showing no armor on the head hitbox and yellow internals. This because I hit the target with the 2 PPCs on the right and not those to the left, because HSR rolled me back to a previous position, where the PPCs on my left were still into cover and could not hit the enemy head. There I understood.
The trouble is that this is an online game. If I'm quicker (minus the ping) I should be able to hit and kill things with no hassle.
Now you need to time aiming and movement, wait few instants before shooting or staying out of the cover for more time than what it should be necessary, and the result isn't pretty good from a competitive pov. At all.
Edited by John MatriX82, 18 June 2013 - 03:32 AM.
#38
Posted 18 June 2013 - 04:13 AM
Skadi, on 15 June 2013 - 02:01 PM, said:
I already do this, just with actual spiders.
Quote
If you get out of cover, as soon as you are out you shoot (moving the crosshair on the target AND as soon as it's on it, click to unleash the weapons) you'll see no hit. This because HSR rewinds your shooting position, few instants before the position that you saw in your client (let's say, your ping + some other time), this makes the server think you shot the cover in front of you, since HSR rolls you back when you were still getting the clearance for your weapons.
The same happens when aiming, if you make last-time adjustments to a target; let's say you aim at the CT, but as soon as the paperdoll updates you see that the RT is critical and you swing your crosshair to hit the RT, notice how the damage is dealt to the CT and not the RT that has been graphically hit. This because HSR rolls your aiming position and physical mech posit...
That's...not what HSR does. What lag is, fundamentally, is one of the machines involved, the client or the server, lagging behind the other, so one machine's present is the other's past. This is what you just described--the server thinks you're still behind cover when your client shows you out in the open.
What HSR does--literally the entire point--is track track how far your client is behind the server. When you shoot, it rolls the server's history of events back by that amount to see where the mechs appeared to be to you when you fired your weapons. The 'H' in HSR stands for "Host."
HSR does not roll your simulation back to see where the server thinks you were when you shot. That would require your client to be ahead of the server--ahead of the machine that's dictating the timeline of the simulation. That's not possible, even if it's a peer-to-peer setup and your machine is functioning as the host.
The problem here is that it's doing HSR for everyone. When you take a shot, the server has to compensate for your lag, but it also has to compensate for the lag of the person you're shooting at. It's entirely possible for none of the involved computers--shooter, target, and server--too agree on where everyone is. There are a variety of way of to deal with this, but ultimately it's impossible (not just hard, but literally impossible) to deal with lag perfectly 100% of the time. Some shots that shouldn't register will, and some that should won't.
It's entirely possible that there's something wrong with PGI's implementation of HSR, or that their approach to dealing with no-win, unreconcilable lag is a poor one. However, we don't have anywhere near enough information to determine that.
Koniving, on 17 June 2013 - 12:07 PM, said:
When an AC/20 is detected to hit between the side and center torsos with no favoring between one way or the other, the AC/20 does zero damage. It has less to do with detection and more to do with the person aiming. This is especially noticeable against Spiders because, simply put, they're the easiest ones to hit in between the CT and STs.
This sounds kinda fishy, but it'd also be very PGI to have it do no damager rather than picking at random or oh I dunno evenly dividing between both. Hell, it could even be an attempt to reduce the impact of splash damage.
Edited by Blue Footed Booby, 18 June 2013 - 04:21 AM.
#39
Posted 18 June 2013 - 06:21 AM
Haitchpeasauce, on 17 June 2013 - 08:38 PM, said:
Or something you've come across in your testing?
I should spend some time in the testing grounds ...
Not documented.
Something I observed. I was with someone, whose kdr is obscene and ELO also pretty high.
He kept complaining about how he would hit light mechs and they wouldn't register damage. And since whenever I play with him every single possible player there is has a PPC, a gauss rifle, two AC/20s, or some combination thereof in multiples (twin Gauss + PPC, triple PPC + gauss, etc.) it didn't long for me to wind up dead. In fact it took seconds of stepping away from the base to get headshot.
So since I kept dying early I started to observe him when he would start fighting lights. He kills anything else with ease.
He'd go for the legs of this Jenner and clearly miss. Boom. Miss. Literally the shots would go on either side of the target. Boom. Miss. He'd lead sometimes and peg an arm with only one AC/20. "Both hit that arm and he still has it!" In this case the Jenner is close and his shots are not hitting exactly where the circle is. That's because AC/20s don't converge well when leading your target since the shot is lined up to go at the ground 130 meters away, not to hit a target that's 80 meters in front of you.
In a different fight, a Spider was running directly at him. No leading, no convergence issues. The only thing is the target is rapidly getting close while going in a straight line. AC/20s let loose. Since the target is closer to him than at the moment he fired, the bullets are trying to converge at the point past the Spider. As such they did not hit center torso, but strangely enough to hit the lines right between the center and side torsos. Nothing registered.
This really got my attention. So I began recording as when I record the speed of the game cuts down to 30 frames per second; this allows me -- during gameplay -- to see finer details such as the shape of bullets.
When he fired at another point in time, I saw one AC/20 hit that space in between and the other hit a side torso. The one that hit the side torso's edge did lots of damage. The one that hit the space in between center and side did nothing at all.
Now, mind you there's always the chance of a ping delay. His ping is usually in the early 50's, mine is typically in the 60's and 70's. We did see different things, though his perspective is likely a combination of both wishful thinking (as one time he declared a hit, I clearly saw both shots go under the spider's armpit) and in one case 'poptarting' with his jagermech from behind cover where I asked him, "Why did you shoot before you got up?" "I didn't. I shot as soon as I peeked."
After observing, I made my first AC/20 Jager out of the JM6-A (my least-liked jager) and then set about
That's not to say I didn't have misses. That still happened. But when I hit it typically registers.
A few matches later I found a spider near a base that had disconnected at the start of the match. I fired at close range, crosshair on the CT. Close as in I'd be inside it if I were much closer. AC/20s fired, and I hit left and right arms. Stepped back. Fired. Left and right torso. Stepped back, fired. No register. Shots when between the torsos. They exploded, but no damage. Stepped back, fired. Cored the CT with both shots hitting it.
AC/20 catapults will have an easier time hitting spiders (their convergence is less affected at close ranges).
If you fire one AC/20 at a time at a light, you waste less if you miss, and you hit where you expect when you do. I have left cannon (bottom) on my left mouse. Right cannon (top) on my right mouse. And then both under a third mouse button under my left finger.
Back in the day AC/20s used to have a splash damage-like effect where you did some damage even if you didn't hit directly. I miss that.
Lots of things could use splash damage, but instead of bonus damage on top of the impact, it should be a division of the damage as a whole. It does 20 damage? Welp, it splashes onto two torsos. 10 damage each totalling 20.
(Edit: "Doing testing this" didn't sound right.)
Edited by Koniving, 18 June 2013 - 06:26 AM.
#40
Posted 18 June 2013 - 07:53 AM
Observations:
Lights will come to a complete stop and then fire. I do not know if this is just an appearance thing, a server thing, or lights secretly taking advantage of some damage soaking secret - perhaps a simple breathing technique? Or maybe this is just what light pilots do now.
I noticed that lights are taking reduced damage from artemis srm6s hitting them point blank, while they are stationary. The result is maybe some scarring, about 3 vollies may get them to orange in one location. More HSR problems?
4 user(s) are reading this topic
0 members, 4 guests, 0 anonymous users